1、Designation: E 1790 04Standard Practice forNear Infrared Qualitative Analysis1This standard is issued under the fixed designation E 1790; the number immediately following the designation indicates the year oforiginal adoption or, in the case of revision, the year of last revision. A number in parent
2、heses indicates the year of last reapproval. Asuperscript epsilon (e) indicates an editorial change since the last revision or reapproval.1. Scope1.1 This practice covers the use of near-infrared (NIR)spectroscopy for the qualitative analysis of liquids and solids.The practice is written under the a
3、ssumption that most NIRqualitative analyses will be performed with instruments de-signed specifically for this region and equipped with comput-erized data handling algorithms. In principle, however, thepractice also applies to work with liquid samples usinginstruments designed for operation over the
4、 ultraviolet (UV),visible, and mid-infrared (IR) regions if suitable data handlingcapabilities are available. Many Fourier Transform Infrared(FTIR) (normally considered mid-IR instruments) have NIRcapability, or at least extended-range beamsplitters that allowoperation to 1.2 m; this practice also a
5、pplies to data fromthese instruments.1.2 This standard does not purport to address all of thesafety concerns, if any, associated with its use. It is theresponsibility of the user of this standard to establish appro-priate safety and health practices and determine the applica-bility of regulatory lim
6、itations prior to use.2. Referenced Documents2.1 ASTM Standards:2E 131 Terminology Relating to Molecular SpectroscopyE 1252 Practice for General Techniques for Obtaining In-frared Spectra for Qualitative AnalysisE 1655 Practices for Infrared, Multivariate, QuantitativeAnalysis3. Terminology3.1 Defin
7、itionsFor definitions of general terms and sym-bols pertaining to NIR spectroscopy and statistical computa-tions, refer to Terminology E 131.3.2 Definitions of Terms Specific to This Standard:3.2.1 interactance, nthe phenomenon whereby radiantenergy entering the surface of a material is scattered by
8、 thematerial back to the surface, but at a different portion of thesurface.3.2.1.1 DiscussionThis differs from diffuse reflectance,where the returning radiation exits the same portion of thesurface of the material as the illuminating radiation entered.3.2.2 training sample (otherwise called a “refer
9、encesample” or “standard”), na quantity of material of knowncomposition or properties, or both, presented to an instrumentfor measurement in order to find relationships between themeasurements and the composition or properties, or both, ofthe sample.3.2.2.1 DiscussionThis term is typically used in c
10、onjunc-tion with computerized methods for ascertaining the relation-ships.Training samples for quantitative analysis (also called “calibrationsamples,” as in Practices E 1655) have different requirements thantraining samples used for qualitative analysis.4. Significance and Use4.1 NIR spectroscopy i
11、s a widely used technique for quan-titative analysis, and it is also becoming more widely used forthe identification of organic materials, that is, qualitativeanalysis. In general, however, the concept of qualitative analy-sis as used in the NIR spectral region differs from that used inthe mid-IR sp
12、ectral region in that NIR qualitative analysisrefers to the process of automated comparison of the spectra ofunknown materials to the spectra of known materials in orderto identify the unknown. This approach constitutes a librarysearch method in which each user generates his own library.4.2 Historic
13、ally, NIR spectroscopy as practiced with classi-cal UV-VIS-NIR instruments using methods similar to thosedescribed in Practice E 1252 was not considered to be a strongtechnique for qualitative analysis. Although the positions andintensities of absorption bands in specific wavelength rangeswere used
14、to confirm the presence of certain functional groups,the spectra were not considered to be specific enough to allowunequivocal identification of unknown materials. A few impor-tant libraries of NIR spectra were developed for qualitativepurposes, but the lack of suitable data handling facilities1This
15、 practice is under the jurisdiction of ASTM Committee E13 on MolecularSpectroscopy and Chromatography and is the direct responsibility of SubcommitteeE13.11 on Chemometrics.Current edition approved Nov. 1, 2004. Published December 2004. Originallyapproved in 1996. Last previous edition approved in 2
16、000 as E 1790 00.2For referenced ASTM standards, visit the ASTM website, www.astm.org, orcontact ASTM Customer Service at serviceastm.org. For Annual Book of ASTMStandards volume information, refer to the standards Document Summary page onthe ASTM website.1Copyright ASTM International, 100 Barr Harb
17、or Drive, PO Box C700, West Conshohocken, PA 19428-2959, United States.limited the scope of qualitative analysis severely. Furthermore,earlier work was limited almost entirely to liquid samples.4.3 Currently, the mid-IR procedure of deducing the struc-ture of an unknown material via analysis of the
18、locations,strengths, and positional shifts of individual absorption bandsis generally not used in the NIR.4.4 With the development of specialized NIR instrumentsand mathematical algorithms for treating the data, it becamepossible to obtain a wealth of information from NIR spectrathat had hitherto go
19、ne unused. While the mathematical algo-rithms described in this practice can be applied to spectral datain any region, this practice describes their application to theNIR.4.5 The application of NIR spectroscopy to qualitativeanalysis in the manner described is relatively new, and proce-dures for thi
20、s application are still evolving. The application ofchemometric methods to spectroscopy has limitations, and thelimitations are not all defined yet since the techniques arerelatively new. One area of concern to some scientists is theeffect of low-level contaminants. Any analytical methodologyhas its
21、 detection limits, and NIR is no different in this regard,but neither would we expect it to be any worse. Since therelatively broad character of NIR bands makes it unlikely thata contaminant would not overlap any of the measured wave-lengths, the question would only be one of degree: whether agiven
22、amount of contaminant could be detected. The user mustbe aware of the probable contaminants he is liable to run intoand account for the possibility of this occurring, perhaps byincluding deliberately contaminated samples in the training set.5. General5.1 NIR qualitative analysis is conducted by comp
23、arison ofNIR absorption spectra of unknown materials with those ofknown reference materials. Since the absorption bands of manysubstances of interest are less distinctive in the NIR than in themid-IR spectral region, the analytical capability of the tech-nique relies heavily on the accuracy of the a
24、bsorption mea-surements and the relationship of the relative absorbances atdifferent wavelengths. Materials to be identified are measuredby a NIR spectrometer, and the spectral data thus generated aresaved in an auxiliary computer attached to the spectrometerproper. One of the several algorithms des
25、cribed in Section 6 isthen applied to the data in order to generate classificationcriteria, which can then be applied to data from unknownsamples in order to classify (or identify) them as being thesame as one of the previously seen materials. Good chemicallaboratory practice should be followed to h
26、elp ensure repro-ducible results for each material. The preparation and presen-tation of samples to the instrument should be consistent withina library, and unknowns should be treated the same way thatthe training samples were.5.1.1 The technique is applicable to liquids, solids, andgases. For analy
27、sis of gases, multipath vapor cells capable ofachieving up to 100-metre path lengths may be required.Spectra of vapors and gases may be sensitive to the totalsample pressure, and this has to be determined for each type ofsample.5.1.2 Unknown samples to be identified may be prescreenedbased on criter
28、ia other than their NIR spectra (for example,visual inspection). The training samples (that is, the “knowns”used to teach the algorithm what different materials look like)may also be similarly prescreened and grouped into libraries ofsimilar materials (for example, liquids and solids). The un-known
29、is then compared with only those materials in theappropriate library. The prescreening will help reduce thechance of false identification, although care must be taken thatan unknown material not in the library is not identified as asimilar material that is in the library.5.1.3 Measurements may be ma
30、de via transmission, reflec-tion, or any other optical setup suitable for collecting NIRspectra. In practice, only transmission and diffuse reflectionhave been in common use.5.1.4 Determination of the relationships between absor-bances at different wavelengths for a set of materials andconsolidation
31、 of these relationships into a set of criteria foridentifying those materials requires the use of computerizedlearning algorithms. These algorithms can also take intoaccount extraneous variations such as are found, for example,when measurements are made on powdered solids.5.1.5 Instrumentation is co
32、mmercially available for makingsuitable measurements in the NIR spectral region. Manufac-turers instructions should be followed to ensure correctoperation, optimum accuracy, and safety before collecting data.5.1.6 NIR spectroscopy has, as one of its paradigms, thatlittle or no sample preparation be
33、required. In conformancewith that paradigm, sample preparation steps in other spectro-scopic technologies are replaced with sample presentationmethodologies in NIR analysis. The most common samplepresentation methods are the following:5.1.6.1 Diffuse ReflectanceSolid materials are ground intopowder
34、(or used as-is, if already in suitably fine powder form)and packed into a cup, which allows the surface of the sampleto be illuminated and the reflected radiant power measured.5.1.6.2 “Transflectance”Clear or scattering liquids areplaced in a cup containing a transparent window with adiffusely refle
35、cting material behind the sample. Any radiantenergy passing through the sample is reflected diffusely by thebacking material, so the net measurement is just like the diffusereflectance measurement of powdered solids.5.1.6.3 TransmissionLiquids or solids are placed in cellswith two transparent window
36、s and measured by transmission.5.1.6.4 Fiber ProbesIlluminating and collecting fibers arebrought in parallel to the sample. A variety of optical configu-rations are used to couple the radiant energy from the fibers tothe sample and back again, in an optical “head” of some sort.Transmittance, reflect
37、ance, and interactance have all been usedat the sample end of the fiber to couple the radiation to thesample. Interactance measurements are sometimes made by thesimple expedient of pressing the end of a fiber bundlecontaining mixed illuminating and receiving fibers against thesample surface.5.2 To c
38、onnect the mathematics with the spectroscopy used,the procedure can be generally described as follows:(1) The spectral measurements define some multidimen-sional space. The axes in that space are the absorbances at thevarious wavelengths, or some mathematical transformationthereof.E1790042(2) Groups
39、 of spectra for the same material define someregion in the multidimensional space.(3) The analysis involves determining which region theunknown falls in.5.2.1 Problems with this type of analysis include the fol-lowing: insufficient separation of the groups in the multidimen-sional space to allow for
40、 classification (indicating insufficientdifferences among the spectra of the materials involved),inadequate representation of measurement variability withingroups during training (indicating an insufficient number orvariety of training samples), or poor detection limits for minorcontaminants.5.2.2 T
41、o optimize the methods against these potential prob-lem areas, generation of a method occurs in three stages. In thefirst, or training stage, known samples are presented to theinstrument. The data collected are then presented to one of thevarious algorithms and are thus used to “train” the algorithm
42、 torecognize the various different materials.5.2.3 In the second, or validation stage, the ability of thealgorithm to correctly recognize materials not in the trainingset of samples is tested. Samples measured during the valida-tion stage should preferably be in the same phase and physicalcondition
43、as the known samples were during the training stage.5.2.4 In the third, or use stage, unknown samples arepresented to the instrument, which then compares the data soobtained to the data from the known samples and decideswhether the data from the unknown agrees with the data fromany of the known mate
44、rials. The unknown material is classifiedas whichever material gives the closest agreement to the data.5.2.5 Optionally, the algorithm may provide for the case inwhich the data from the unknown does not agree with that fromany of the knowns sufficiently well to permit identification, andrefuse to id
45、entify the unknown sample.5.3 Samples to be identified during the use stage must be inthe same phase and physical condition as the known sampleswere during the validation stage.5.3.1 Liquids may be run neat or in solution. In either case,the optical pathlength of the sample cell should be fixed, be
46、thesame for all liquids to be compared with a given unknown, andbe specified as part of the method. While an algorithm may betrained on data incorporating variations in these characteristics,greater accuracy will be achieved when extraneous variationsare reduced. The unknown, of course, should also
47、be run in acell under the same conditions as the training samples. If asolution is used, the amount of dilution should also bespecified.5.3.2 Some solids may be run as-is if they have one or moresuitably flat surfaces; others may need to be ground. If solidsamples are ground, the same procedure shou
48、ld be used for allmaterials in a given library, and that procedure should bespecified as part of the method.5.3.3 The unknowns must also be treated in the samemanner as the training samples. It is particularly important thatif the samples must be ground, the unknown samples should beground to the sa
49、me particle size as the known samples includedin the library.6. Algorithms Used6.1 This section describes some of the computerized algo-rithms that have been found effective for qualitative analysis inthe NIR spectral region. This section is mainly for reference.Descriptions of multivariate methods of statistical data analysistend to be inherently abstract mathematically and resistant toreduction to words. A number of books exist in both thestatistical and chemometric literature that describe methods ofmultivariate analysis at varying levels of mathematical abstrac-tion (se
